README.md

---
license: apache-2.0
tags:
- generated_from_trainer
base_model: bert-base-uncased
datasets:
- conll2003
metrics:
- precision
- recall
- f1
- accuracy
model-index:
- name: NER_Pittsburgh_TAA
  results:
  - task:
      type: token-classification
      name: Token Classification
    dataset:
      name: conll2003
      type: conll2003
      config: conll2003
      split: validation
      args: conll2003
    metrics:
    - type: precision
      value: 0.9429236395877203
      name: Precision
    - type: recall
      value: 0.9517843159190066
      name: Recall
    - type: f1
      value: 0.9473332591025497
      name: F1
    - type: accuracy
      value: 0.9867030994328562
      name: Accuracy
language:
- en
- uk
---

<!-- This model card has been generated automatically according to the information the Trainer had access to. You
should probably proofread and complete it, then remove this comment. -->

# NER_Pittsburgh_TAA

This model is a fine-tuned version of [bert-base-uncased](https://huggingface.co/bert-base-uncased) on the conll2003 dataset.
It achieves the following results on the evaluation set:
- Loss: 0.0860
- Precision: 0.9429
- Recall: 0.9518
- F1: 0.9473
- Accuracy: 0.9867

## Model description

## Ukr

Модель була створена як практичне завдання з машиного навчання, це за fine-tuning BERT модель для задачі Named Entity Recognition. 
Датасет який був використан це conll2003, стандат для навчання моделей під задачу Named Entity Recognition, або ще визначення складових мови в реченні.
Дізнатися як працює модель маєте змогу або через інтерфейс, який надає huggingface, або ж через код 

from transformers import AutoTokenizer, AutoModelForTokenClassification

tokenizer = AutoTokenizer.from_pretrained("CineAI/NER_Pittsburgh_TAA")

model = AutoModelForTokenClassification.from_pretrained("CineAI/NER_Pittsburgh_TAA")

Якщо цікавить чому модель має таку назву, перше це для чого вона для NER, друга складова це назва крутої пісні Pittsburgh третя і остання складова
це гурт який пісню створив це The Amity Affliction

## En

The model was created as a practical machine learning task, it is a fine-tuning BERT model for the Named Entity Recognition task. 
The dataset used is conll2003, a standard for training models for the Named Entity Recognition task, or for identifying the components of speech in a sentence.
You can find out how the model works either through the interface provided by huggingface or through the code 

from transformers import AutoTokenizer, AutoModelForTokenClassification

tokenizer = AutoTokenizer.from_pretrained("CineAI/NER_Pittsburgh_TAA")

model = AutoModelForTokenClassification.from_pretrained("CineAI/NER_Pittsburgh_TAA")

If you are wondering why the model has such a name, the first is why it is for NER, the second component is the name of a cool song Pittsburgh, the third and last component
is the band that created the song - The Amity Affliction

## Intended uses & limitations

Everyone can use this model, it is completely free and distributed under the Apache 2.0 licence.

## Training and evaluation data

Training and assessment data are the same - conll2003 

## Training procedure

### Training hyperparameters

The following hyperparameters were used during training:
- learning_rate: 2e-05
- train_batch_size: 32
- eval_batch_size: 32
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 5

### Training results

| Training Loss | Epoch | Step | Validation Loss | Precision | Recall | F1     | Accuracy |
|:-------------:|:-----:|:----:|:---------------:|:---------:|:------:|:------:|:--------:|
| No log        | 1.0   | 439  | 0.0863          | 0.9437    | 0.9444 | 0.9440 | 0.9861   |
| 0.0024        | 2.0   | 878  | 0.0995          | 0.9394    | 0.9442 | 0.9418 | 0.9852   |
| 0.0021        | 3.0   | 1317 | 0.0904          | 0.9355    | 0.9463 | 0.9409 | 0.9856   |
| 0.0012        | 4.0   | 1756 | 0.0835          | 0.9427    | 0.9514 | 0.9471 | 0.9867   |
| 0.0009        | 5.0   | 2195 | 0.0860          | 0.9429    | 0.9518 | 0.9473 | 0.9867   |


### Framework versions

- Transformers 4.41.2
- Pytorch 2.3.0+cu121
- Datasets 2.19.2
- Tokenizers 0.19.1